Research On Key Problems Of Face Detection Recognition And Tracking Technology

Research on Key Problems of Face Detection Recognition and Tracking TechnologyFace detection means to adopt a certain strategy to search any set of given images in order to determine whether they contain human faces. It will return to the location, size and expression of one's face if so. Face recognition refers to a computer technology that carries out identity authentication by means of using analysis and comparison of information of visual characteristics of human faces. Target tracking refers to parameters that image sequences identify targets and meanwhile position them precisely to get access to the target movements, such as position, velocity; acceleration and trajectory and so on so that they can be processed and analyzed further to understand the behaviors of moving targets and to complete higher-level visual tasks.Face detection is a complex and challenging problem of pattern detection, whose main difficulties rest in two ways:partly caused by inner changes in the faces-faces may have a rather complex change in details, different appearances, such as face shapes and colors etc; different expressions, such as opening and closing of eyes and mouths etc.; blocking of human faces, such as glasses, hair and head accessories as well as other external objects and so on. On the other hand due to changes caused by external conditions:different multi-expressions caused by different imaging angles, such as plane rotation, deep rotation and top and bottom spin, in which deep rotation has a greater impact; Lighting effects, such as image brightness, contrast changes and shadows and so on; Imaging conditions of images, such as focal length, imaging distance and access approaches of images of photography equipments and so on.Generally a face recognition system consists of image capture, face positioning, image pre-processing as well as face recognition (identification or identity search). System input generally contains one or a series of undetermined human face images as well as a number of face images with known identity in the database or the corresponding codes while its output refers to a series of similarity scores, indicating identity of faces needed to be recognized.To resolve problems of face detection, there are some difficulties occurring. If you can find a number of related algorithms and can achieve real-time effects in the application process, you will provide guarantee for face detection recognition and the tracking system with successful structures and practical application values.Face detection and recognition and tracking technology is a very important research direction in the field of computer vision studying with the combination of a variety of knowledge in multidisciplinary field as one of the research content and develops fast with popularization and applications of information construction and internet.The central theme of this paper is to study the key issues in the field of face detection and recognition and tracking with various problems and difficulties to improve as the objective of the study. A lot of analysis on history and current situation of face detection and recognition and tracking technology is made and a variety of popular techniques, methods and ideas in the related fields have been fully understood through this analysis making people more familiar with performance differences and features displayed in the practical application of these technologies. On the basis, a more in-depth and thorough study has been made to get a series of improvements in many key issues in the field of face detection and recognition and tracking technology.Detailed breakthrough and innovation is specifically in the following aspects.Better methods for improvements in face detection and recognition have been proposed:Although the use of support vector machines for face detection and recognition has good accuracy and comparative effects, the good performance can be displayed only when the sample size is large enough. At this time efficiency of the system gets lower and demands shorter computing time thus posing a heavy burden to the hardware system.So,Starting from improving speed and accuracy of real-time detection, this paper proposes the face detection algorithm combined fast template matching and SVM. There is great improvement in speed and meanwhile it maintains the advantages of anti-interference of the SVM classifiers and strong stability.For some problems of face tracking in, dynamic video detecting, the improved tracking algorithm has been proposed:1. CamShift algorithm is obviously good in the speed of face detection and can complete the scanning process very fast. It is based on the matching algorithm of color information to look for color areas for matching with the target in a series of dynamic image sequences and is prone to lose in the event of skin color blocking and interference. So, Starting from adapting to the objective and required environment for blocking, it improves loss in Camshift as a result of blocking in tracking and meanwhile maintains the advantages of the fast original algorithm.2. Particle filter algorithm has good resistance to environmental interference and can get a good tracking effect even if the change in application environment is very complicated. It is a tracking process that simulates the target state based on a large number of sampling particles. The more the sampling particles, the more accurate the algorithm is and the higher the corresponding computational complexity of the algorithm.For the robustness of particle filter algorithm and real-time of Mean-Shift, a self-adaptive tracking method that combines two kinds of tracking frameworks, combining the above two advantages organically is proposed and it greatly increases the speed under the premise of not losing reliability.3. Through the automatic tracking and identification of the specific state of the eyes in human-computer interaction systems, it can complete the process of special instructions and is also applied concretely to similar driver fatigue detection and so on.For requirements for human eyes tracking and state recognition in human computer interaction systems, a highly efficient tracking and identification method is proposed so that it can finish the task of eyes tracking and state recognition with less resource consumption well.This paper fully describes popular techniques, methods, frameworks and development trends in the area of face detection, identification and tracking and gets a lot of research results combining acquired theories, techniques, methods and experiences. The achievements enrich the research contents in the field above and have a certain theoretical significance and reference value.